Competitor research for a US SaaS entering Japan: a 90-day plan

What you are trying to learn in 90 days

You are answering three questions, in order: who sells what in Japan and under which legal entity, how Japanese buyers find and compare those products, and which claims about them you may print. Everything else, including win rates, quoted prices and partner arrangements, is not public, and the plan says so rather than guessing.

The English habits you bring do not transfer cleanly. G2 and Crunchbase may hold little on Japanese incumbents (check before relying on them), and some Japanese vendors take price inquiries instead of publishing a price. The sources below replace them. The method for searching in Japanese with Claude, including how the search locale and the output language are separate choices, is covered in Japanese competitor sources; this plan assumes it and adds the sequence.

Keep the work in a Claude Project set up as in the project setup lesson, with a competitor sheet and a source log in it from day one. Claude Research needs a paid plan with web search on and returns citations you can check, per Anthropic's Research help page (updated 2 June 2026). Its web search page says location is inferred from your IP address, so a Japanese query from a US office can still return US-leaning results. State the output language in the prompt and write Japanese terms in Japanese script.

Days 1 to 14: name the entities

List 8 to 15 candidates in three groups: Japanese incumbents in your category, global rivals with a Japanese presence, and the option of doing nothing, which in Japan often means a spreadsheet and an internal team. Ask Claude Research to propose candidates from Japanese queries, then confirm each one yourself.

1. Tie every name to a registered entity

   The National Tax Agency's corporate number site publishes three basic items for every registered company: trade name, head office address and corporate number, with download and Web API access. Search it for each candidate. A name Claude cannot tie to a corporate number, or to a stated Japanese subsidiary on the company's own 会社概要 page, comes off the list. Name collisions are common: ITreview's product page for a "CRAYON" is a Japanese customer support tool from CRAYON 株式会社, not the US competitive intelligence vendor.

2. Pull capital, headcount and entity type from gBizINFO

   gBizINFO, run by METI, holds government data on more than five million companies. Filters include capital, employee count and founding year, and the entity type filter has a category for foreign companies, so a US rival's Japanese arm is findable. Fields are present only where the company has filed them; an empty cell is unknown, not zero.

3. Book the JETRO consultation if your firm qualifies

   JETRO's support page states that its staff and experts support foreign and foreign-affiliated companies planning to establish or expand in Japan free of charge, partially for a fee, including individual consultations on the market, industries and regulations. Check JETRO's eligibility terms before you book. Ask what sector research JETRO already holds before you buy anything.

This prompt's wording forces the entity check:

Search in Japanese for companies selling [category] to Japanese B2B buyers.
For each candidate return: the company name in Japanese, its corporate
number if you can find it on the National Tax Agency site or the company's
会社概要 page, the product name, and the URL where you saw it. Mark any
company you could not tie to a registered entity as "unconfirmed". Do not
summarise the market; list candidates only.

Days 15 to 45: read what listed companies must publish

A listed Japanese competitor publishes more than its marketing site ever will. Three systems hold it, and each has a limit.

1. EDINET for securities reports

   EDINET is the Financial Services Agency's electronic disclosure system for securities reports (有価証券報告書) and related filings under the Financial Instruments and Exchange Act. Its June 2026 guide says anyone can view filings freely over the internet and download the XBRL data, and adds that the English information contained in that XBRL data is not guaranteed to be accurate. Read the Japanese; the segment table, employee count and risk factors are the sections a PMM uses.

2. TDnet for timely disclosures, with a 31-day window

   The Tokyo Stock Exchange's TDnet page states that disclosures stay on the free viewing service for 31 days including the disclosure date, that ten years of past disclosures are available through the listed company information service, and that five years of complete data are in a paid database. The page itself is dated 28 February 2022, so recheck it. A disclosure you missed within 31 days needs the ten-year archive, so set a monthly reminder.

3. PR TIMES for what the company chose to say

   PR TIMES reported in its first-quarter results release of 14 July 2026 that 129,233 companies used the service at the end of May 2026 and that 65.8% of domestically listed companies are on it. By the same figure, about a third of listed companies are not, and a release is a company-chosen channel: no release does not mean no event. Reading a competitor's feed in order, and filing each release against your sheet, is covered in monitoring competitors on PR TIMES.

Private competitors and Japanese subsidiaries of foreign firms file none of this. For them, the Tokyo Metropolitan Library's business information list (September 2024) names the printed directories: Tokyo Shoko Research's 東商信用録, roughly 270,000 companies with an A, B or C rating, Teikoku Databank's yearbook of about 140,000, and the unlisted-company 会社四季報 of about 13,000. Teikoku Databank's lineup page sells credit reports through a membership and lists per-sheet prices (for example 5 inquiry sheets for 120,000 yen as of 12 October 2026), alongside a company database; Tokyo Shoko Research's public service page shows no prices. Buy one report for your top rival if you must, and write "unknown" in the revenue cell for everyone else.

Days 46 to 75: see the market as a Japanese buyer does

ITreview and BOXIL are the Japanese review sites. Map the categories a buyer in your space would browse, then read the reviews there; the method is in reading ITreview and BOXIL reviews. The table sizes what you will find.

Read job postings one page at a time and cite the URL. Wantedly's terms (revised 23 April 2026) prohibit crawling, scraping and data mining without prior consent, by bot, AI agent or hand. Indeed Japan's terms (updated 30 September 2026) limit a job seeker to personal, non-commercial job searching. We could not reach doda's terms. Check each site's current terms with your legal team. No source we opened shows that a hiring pattern predicts a product move; record the inference as a hypothesis.

What is not public

Add two gaps to the list. Claude web search can miss PDFs and login-gated pages, so a filing you know exists may need a manual download. And no Japanese-built battlecard or competitive intelligence platform turned up in the searches we ran; the closest products are company databases such as Speeda, which claims more than 2,500 customers, INITIAL for startup information, and SalesNow. Record that as "not found"; the searches do not show that none exists.

Days 76 to 90: checking a comparison before you publish it

The Consumer Affairs Agency's comparative advertising page states that comparison with a competitor's goods or services is not itself prohibited or restricted, and lists three requirements a comparison should satisfy: the content claimed is objectively substantiated, the substantiated figures and facts are quoted accurately and appropriately, and the method of comparison is fair. The same page adds that a study by a third party unrelated to the advertiser is considered objective, and that an advertisement citing a study should show the research body, the date and the place of the research.

Two fairness traps on that page map directly to software comparisons. Emphasising a feature that barely affects the product's overall function as if the whole product were superior may be a misleading representation, according to the agency. So may stating an advantage while omitting, or failing to state clearly, a disadvantage inseparable from it. The agency's own example is a cram school that claimed the top university pass rate using a counting method other schools did not share.

• Every claim on the slide traces to a row in the ledger with a source URL, a quote in Japanese, a date and a scope.
• Every number came from a named third party or from the competitor's own filing, with the research body, date and method shown next to it.
• No feature is presented as decisive when it is minor to the product as a whole.
• Every stated advantage names the disadvantage that comes with it.
• No ranking or 'No. 1' claim rests on a counting method the other party did not use.
• A Japanese lawyer or native reviewer has read the final copy.

This section has two limits. The statute is framed around misleading the general consumer, and whether it reaches advertising aimed only at businesses was not resolved by any source we read, so do not assume a B2B exemption. Separately, the Unfair Competition Prevention Act prohibits announcing or spreading false facts that harm a competitor's business reputation, per a summary by Monolith Law Office (updated 5 January 2025); sources disagree on the item number, and this lesson is not legal advice. Have a Japanese lawyer review any comparison before you publish it.

Keeping it current, and where Steve fits

On day 90 the sheet is accurate. By day 120 it is not: a disclosure has passed its 31-day window, a competitor has posted on PR TIMES, and a new review has landed. The DIY remedy is a monthly rerun of phases two and three against the same ledger, with the source log telling you which pages to reopen.

Steve covers part of that upkeep. Its competitive intelligence page describes daily competitor news summaries sent to the channels you choose, up to 25 custom news sources per project refreshed daily, and battlecards where each row shows Confirmed, Needs verification or Unverified with links to sources; updates wait for your review. That matches the ledger discipline above, for the news and website side of the sheet.

The fit stops there, and the site is plain about it. Steve does not claim to read EDINET, TDnet, credit-bureau reports or job postings, does not name PR TIMES, ITreview or BOXIL among its sources, and does not claim Japanese-language sources or a Japanese product interface; it has a Japanese website. For a one-off entry study, for listed-company filings and for any legal check, the plan above is the tool. If keeping the news and battlecard side current after launch is the step your team drops, that is the page to read.